linux-kernel - Re: [PATCH v10 02/13] x86/insn-eval: Compute linear address in several utility functions

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20171104004008.GA27257@voyager>
Date:   Fri, 3 Nov 2017 17:40:08 -0700
From:   Ricardo Neri <ricardo.neri-calderon@...ux.intel.com>
To:     Ingo Molnar <mingo@...nel.org>
Cc:     Ingo Molnar <mingo@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        "H. Peter Anvin" <hpa@...or.com>,
        Andy Lutomirski <luto@...nel.org>,
        Borislav Petkov <bp@...e.de>,
        Peter Zijlstra <peterz@...radead.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Brian Gerst <brgerst@...il.com>,
        Chris Metcalf <cmetcalf@...lanox.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Paolo Bonzini <pbonzini@...hat.com>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Huang Rui <ray.huang@....com>, Jiri Slaby <jslaby@...e.cz>,
        Jonathan Corbet <corbet@....net>,
        "Michael S. Tsirkin" <mst@...hat.com>,
        Paul Gortmaker <paul.gortmaker@...driver.com>,
        Vlastimil Babka <vbabka@...e.cz>,
        Chen Yucong <slaoub@...il.com>,
        "Ravi V. Shankar" <ravi.v.shankar@...el.com>,
        Shuah Khan <shuah@...nel.org>, linux-kernel@...r.kernel.org,
        x86@...nel.org, ricardo.neri@...el.com,
        Adam Buchbinder <adam.buchbinder@...il.com>,
        Colin Ian King <colin.king@...onical.com>,
        Lorenzo Stoakes <lstoakes@...il.com>,
        Qiaowei Ren <qiaowei.ren@...el.com>,
        Arnaldo Carvalho de Melo <acme@...hat.com>,
        Adrian Hunter <adrian.hunter@...el.com>,
        Kees Cook <keescook@...omium.org>,
        Thomas Garnier <thgarnie@...gle.com>,
        Dmitry Vyukov <dvyukov@...gle.com>
Subject: Re: [PATCH v10 02/13] x86/insn-eval: Compute linear address in
 several utility functions

On Fri, Nov 03, 2017 at 11:17:49AM +0100, Ingo Molnar wrote:
> 
> * Ricardo Neri <ricardo.neri-calderon@...ux.intel.com> wrote:
> 
> > On Thu, Nov 02, 2017 at 09:51:08AM +0100, Ingo Molnar wrote:
> > > 
> > > * Ricardo Neri <ricardo.neri-calderon@...ux.intel.com> wrote:
> > > 
> > > > +	/*
> > > > +	 * -EDOM means that we must ignore the address_offset. In such a case,
> > > > +	 * in 64-bit mode the effective address relative to the RIP of the
> > > > +	 * following instruction.
> > > > +	 */
> > > > +	if (*regoff == -EDOM) {
> > > > +		if (user_64bit_mode(regs))
> > > > +			tmp = (long)regs->ip + insn->length;
> > > > +		else
> > > > +			tmp = 0;
> > > > +	} else if (*regoff < 0) {
> > > > +		return -EINVAL;
> > > > +	} else {
> > > > +		tmp = (long)regs_get_register(regs, *regoff);
> > > > +	}
> > > 
> > > > +	else
> > > > +		indx = (long)regs_get_register(regs, indx_offset);
> > > 
> > > This and subsequent patches include a disgustly insane amount of type casts - why?
> > > 
> > > For example here 'tmp' is 'long', while regs_get_register() returns
> > > 'unsigned long', but no type cast is necessary for that.
> > > 
> > > > +			ret = get_eff_addr_modrm(insn, regs, &addr_offset,
> > > > +						 &eff_addr);
> > 
> > One of the goals of this series is to have the ability to compute 16-bit, 32-bit
> > and 64-bit addresses. I put lost of casts, between signed and unsigned types,
> > between 64-bit and 32-bit and 16-bit casts. After seeing your comment I have gone
> > through the code and I have removed most of the casts. Instead I will use masks.
> > I will also inspect the resulting assembly code to make sure the arithmetic is
> > performed in the address size pertinent to each case.
> 
> Well, casts are probably fine when the goal is to zero out high bits

I was able to remove the majority of casts and used masks. I see many other parts
of Linux doing similarly. For instance, in arch/x86/kernel/kexec-bzimage64.c I see

    params->hdr.ramdisk_image = initrd_load_addr & 0xffffffffUL;

ramdisk_image is of type __u32 while initrd_load_addr is of type unsigned long.

I guess that in this example doing

    params->hdr.ramdisk_image = (__u32)(initrd_load_addr & 0xffffffffUL);

would be redundant? The mask would indicate better what is going on.

> but the ones I quoted converted types of the same with.

I made sure I removed these.

> 
> For register values it would also probably be cleaner to use the u8, u16, u32 and 
> u64 types instead of char/short/int/long - this goes hand in hand with how the 
> instructions are documented in the SDMs.

In the rest of the functions I have used char/short/int/long. Would it be OK to have
a mixture of type styles? Perhaps I can rewrite only the functions that deal with
variables of different width.

Plus, one more advantage of using char/short/int/long is that when building a 32-bit
kernel long will be a 32-bit type. Thus, all the aritmetic would be naturally done
with variables of the appropriate width. Perhaps I could use u8/u16/u32/long? It
looks white odd, though.

Thanks and BR,
Ricardo